AbstractSavannah fires are of utmost importance to the livelihoods of rural households living in the miombo woodlands of southeast and southern Africa in carrying out their daily activities. In order to add to the existing knowledge on safe prescribed burning procedures, we developed an experimental physics‐based approach modelled on the implementation of a winter surface fire in northern Mozambique. Twelve experimental fire plots were established and burned to identify local conditions that allow a safe prescribed burn to be conducted in the Miombo woodland, northern Mozambique. Two non‐destructive methods were used to measure the amount of forest fuel, namely: the disk meter method and the triangular line‐intercept sample. The plots were burned alternately and parallel to the prevailing wind direction using a head and backing fire. Relationships between fire behaviour, fuel properties, site characteristics and weather conditions were determined using Analysis of Variance (ANOVA) and Covariance (ANCOVA). The ANOVA models predicted 51% and 55% of fire behaviour variance for the flame height and rate of spread, respectively, while the ANCOVA accounted for 67% and 65% of fire behaviour for the flame height and rate of spread, respectively. Findings from this research suggest that standing fuel with back‐fire produces lower‐intense fire with higher controllability.